starphantom666
starphantom666
> You can import json file with tasks and then export the results in JSON format again. There will be `annotations` field added where the results are stored. Sorry, there's...
> You can import json file with tasks and then export the results in JSON format again. There will be `annotations` field added where the results are stored. In fact,...
> Ok, now it's much clearer. Yes, LS doesn't yet support multiple tasks reading from storages. The similar problem exists with the export. It's a pretty popular request, so I...
Well, such a workflow is also worth trying I have another question to ask, which place is the data stored after the marking of this platform? In fact, when we...
> 是的，我已经完成了，但是发生了一些类似的情况，模型经历了灾难性的遗忘。如果不是特别有必要的事情，我不建议这样做。 但是低资源想要训练领域模型好像只能这样做
(py310) zhangyu@ps:~/project/FastChat/FastChat/fastchat$ python3 -m fastchat.serve.cli --model-pat /home/zhangyu/project/FastChat/FastChat/fastchat/output/vicuna-13b --num-gpus 2 Loading checkpoint shards: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:26 [bsz, 1, tgt_seq_len, src_seq_len] │ │ 462 │ │ combined_attention_mask = None │ │ 463...
> What is your Pytorch version? Could you try the latest pytorch version? > > The command works for me on V100 I also have the same problem. The model...
用p-tuning还是lora做的，lora多卡我都跑不起来，量化直接报错
能不能出个量化后的模型微调